基于小波子带融合的单训练样本掌纹识别方法  

Single training sample palmprint recognition method based on wavelet subbands fusion

作　　者：张延强 李哲谦 王博涵[3] 

机构地区：[1]国家信息中心信息化研究部,北京100045 [2]北京交通大学国际合作交流处,北京100044 [3]国家信息中心公共技术服务部,北京100045

出　　处：《轻工学报》2017年第5期88-94,共7页Journal of Light Industry

基　　金：国家自然科学基金项目(61201158)

摘　　要：针对目前大多数掌纹识别方法对于单训练样本系统识别性能欠佳的问题,提出一种基于小波子带融合的主成分分析方法,用于单训练样本掌纹识别.该方法将小波低频子带与水平和垂直高频子带相结合进行身份识别,使用低通滤波增强相应边缘信息的鲁棒性,以提高高频子带的识别性能,然后采用求和算子对各匹配分数进行融合.实验结果表明,对于单训练样本掌纹识别,该方法平均识别率达89.93%,较传统方法有6%~9%的性能提升.In view of the poor performance of the present most palmprint recognition for single training sample system,a principal components analysis method for single training sample palmprint recognition was presented,which combined multi-subbands of wavelet transformation.This method combined wavelet low frequency subband with horizontal and vertical subbands to identify.Low-pass filter was utilized to enhance the robustness of horizontal and vertical subbands,and the summation operator was used to fuse their matching scores.Experimental results showed that for single training sample palmprint recognition the average recognition rate of the proposed method was 89.93%,which was 6%~9% higher than some of the traditional algorithms.

关 键 词：小波分解 主成分分析 匹配分数融合 单训练样本掌纹识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]